Salesforce Counters AI SaaS Doom

🔑 Enhanced Key Takeaways

• •Salesforce converted nearly 20 trillion tokens into over 2.4 billion AWUs across Agentforce and Slack, with tokens up fivefold year-over-year and AWUs up 57% quarter-over-quarter[4][6].
• •Slackbot AI agent leverages unique business conversation context from Slack, surpassing systems-of-record data, with over 350 AI applications including Anthropic models in its marketplace[6].
• •Salesforce introduced Agent Script, a rule-based scripting layer to enforce predictable AI behavior and address inconsistencies in Agentforce, shifting some responsibility to CIOs for scripting[5].

🛠️ Technical Deep Dive

• •AWU defined as discrete AI agent tasks like updating records, triggering workflows, or making decisions, measured to evaluate token-to-work ratios across models from OpenAI, Anthropic, and others[4][6].
• •Agent Script adds step-by-step rule-based logic to Agentforce for consistent outputs, mitigating inconsistent responses to identical queries[5].
• •Trust Layer in implementations includes consent enforcement, PII masking, reasoning logs, and Instruction Adherence Scores for compliant agent workflows[2].

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline